Breast Cancer Research and Treatment (2006) 97: 285–291
Springer 2006

DOI 10.1007/s10549-005-9122-7

Preclinical study

Human breast duct anatomy, the ‘sick lobe’ hypothesis and intraductal approaches
to breast cancer

James J. Going1 and Timothy J. Mohun2

1
Division of Cancer Sciences and Molecular Pathology, University of Glasgow, Glasgow, Scotland, UK; 2Develop-
mental Biology Division, National Institute for Medical Research, Mill Hill, London, UK

Key words: anatomy, breast, cancer, ductoscopy, human, nipple, pre-cancer

Summary
Introduction. Information about central and peripheral duct anatomy is a requirement for developing intraductal
approaches to human breast cancer, but remains sparse. This study looks at the acquisition and digital modelling of
data describing breast duct branching from thick (‘subgross’) sections using data structures from the neurosciences,
and at high-throughput imaging of duct anatomy in the nipple.
Methods. The branching of a large breast duct system was modelled using data extracted from cleared and
stained 2 mm ‘subgross’ sections of an autopsy breast using a public-domain neuron modelling program (CVAPP),
and episcopic fluorescence image capture (EFIC) was used to collect a stack of 1100 autofluorescence images of
ducts in a mastectomy nipple.
Results. The duct skeleton was captured in 440 line segments with some pruning of terminal ducts. Extracting this
data manually in a usable form was, however, laborious and error prone, emphasising the need for improved
morphological informatics. EFIC captured anatomical detail and subsequent 3D reconstruction was consistent with
the distinction between ‘type A’ and ‘type B’ nipple ducts proposed by Going and Moffat (J Pathol 203: 538–544,
2004).
Conclusions. Whole-lobe duct modelling and EFIC reveal central and peripheral duct anatomy in human breast.
Such knowledge is required for understanding normal breast development, the growth of cancer precursors, and for
developing the intraductal approach to breast cancer.

Introduction central and peripheral duct anatomy need to be filled,

In 1840 Astley Cooper said ‘to dissect and prepare [the
mammary glands’] constituent parts and intimate
structure for a clear demonstration is a very difficult
task; so that I have heard a good anatomist say, ‘‘The
breast is so complicated that I can make nothing clear of
it’’ [1]. This complexity is still challenging: while the
living breast can be imaged by thermography, X ray,
ultrasound and MRI, and tissues and cells can be
examined by histology, electron microscopy and the
techniques of molecular biology, it is difficult for results
of these investigations to be examined in the context of a
complete breast.
Comprehensive 3D mapping of duct branching in a
complete human breast would permit morphological
and molecular investigations to be linked to their ana-
tomical context, illuminating normal breast develop-
ment and the evolution of cancer precursors. With
increasing interest in diagnostic and therapeutic access
to the breast via the nipple (nipple aspiration, duct
lavage, ductoscopy) [2,3], gaps in our knowledge of
but conventional histology is not a viable route to large-
scale breast mapping. To process all of an average 400 g
breast for histology would require more than 200
blocks, and tracing duct branching would require sec-
tions to be examined at multiple levels from every block.
‘Subgross’ methods which derive ultimately from the
work of Leipzig anatomist Werner Spalteholz [4–7] are
more promising. Opaque tissues are made transparent
(‘cleared’) in organic solvents of high refractive index like
methyl salicylate (m=1.53), benzyl alcohol (m=1.54),
benzyl benzoate (m=1.57) or blends like benzyl alcohol/
benzyl benzoate (‘BABB’), and structures stained with
alum carmine or haematoxylin, or blood vessels injected
with India ink, are visible in the cleared tissues. Densely
cellular structures embedded in a sparsely cellular stroma,
like breast ducts and lobules, are seen best and small or-
gans like mouse or rat mammary glands can be studied in
their entirety [8]. Such ‘wholemount’ preparations are
extensively used in experimental studies of rodent mam-
mary gland development, and an entire adult human
breast can be studied in stained and cleared serial sections
286 JJ Going et al.
2–3 mm thick [7]. This paper examines the challenge of
extracting duct branching data from such thick sections of
breast tissue. Breast ducts and neurons are both charac-
teristically branching, so data structures designed to de-
scribe neurons in neurophysiological modelling ought to
be adaptable to describing the breast ducts.
In addition, a novel methodology, episcopic fluores-
cence image capture (EFIC) [9], is evaluated as a tool for
gathering 3D structural information about ducts in the
crowded environment of the nipple. Conventional 3D
reconstruction from images of serial sections has to
overcome two non-trivial challenges: registration
(determining image rotations and translations required
to align each image with the images above and below it
in the image stack) and segmentation (determining which
areas of the image correspond to the object(s) to be
reconstructed). Registration is unnecessary with inher-
ently registered images generated by CT or MRI scan-
ning, confocal or multiphoton microscopy, and EFIC
was devised to generate inherently registered images for
high-throughput structural phenotyping of transgenic
embryos [9]. Tissue embedded in opaque paraffin wax is
sectioned and the cut surface of the tissue remaining in
the block is imaged by autofluorescence after each sec-
tion is cut. Provided no block movement occurs during
section cutting, images are perfectly registered.
Materials and methods
This study pilots the modelling of duct branching in one
mammary lobe of a previously studied autopsy breast
[10,11], and also pilots the application of EFIC to the
generation of detailed 3D structural data from one
mastectomy nipple (not previously studied).
Whole-breast ductal anatomy
Public-domain neuron modelling software CVAPP
(University of Southampton; [12]) was used to model in
3D the largest duct system in a single autopsy breast
previously studied using subgross methods [10,11].
CVAPP treats a neuron as a list of individual line seg-
ments. Each individual segment is represented using this
data structure: <identity number> <diameter>
<type> <parent> <x> <y> <z>. Convention-
ally, the first item in the file is allocated number o. From
this origin a straight line segment <1> extends to ter-
minate at <x1><y1><z1>. If no branching occurs at
this point and the duct does not terminate, line segment
<2><diameter> <type> <1> <x2> <y2>
<z2>, will lead to <3> <diameter> <type> <2>
<x3> <y3> <z3> until a bifurcation is reached at
the end of some line segment k. At this point the data file
follows one branch past several more bifurcations if
need be until a termination is reached. A jump back to
the first bifurcation is made and the next branch, <n>
<diameter> <type> <k> <xk> <yk> <zk> is
followed. Eventually the branching structure is encoded
in its entirety and the file ends.
The <x> <y> and <z> coordinates were ex-
tracted from digital images of duct tracings created for
all 25 breast slices from the original subgross photo-
graphs. All points at which a duct could be observed
entering one side of a thick section, bifurcation points
identifiable within thick sections, and the points at
which a duct was seen to exit from a slice were encoded.
Ducts exiting one slice were almost always identifiable
entering the adjacent thick section, defining a parent/
child relationship between the two duct segments. The
<x> <y> coordinates of the point at which one duct
left the ‘upper’ section did not always match exactly the
<x> <y> coordinates of the point at which same duct
was identified entering the next section, so the <x>
<y> coordinates of the terminus of the ‘parent’ and the
beginning of the ‘child’ were normalised as <(xn+x¢n)/
2> and <(yn+y¢n)/2> where <xn> <yn> and
<x¢n> <y¢n> were the x and y coordinates of the exit
of the ‘parent’ from the slice above and the entry of the
‘child’ into the slice below.
Z coordinates were assigned to each plane of section
starting at zero for the top of slice one, allowing an
arbitrary increment of 10 pixel units for each successive
thick section. The same z measurement was attributed to
the bottom of one section and the top of the next.
Bifurcations within the thickness of a section were
allocated z values by interpolation.
Episcopic fluorescence image capture
An unstained tissue block 2 mm thick containing the
apex of the papilla of a surgical mastectomy breast fixed
in formaldehyde was embedded in paraffin wax dyed a
deep red, which absorbs light emitted from below the
block surface. A total of 1100 images were collected of
serial sections through the tissue at 2 lm intervals. Data
was collected in the original xy plane and in recon-
structions in the xz and yz planes in the laboratory of
Dr Tim Mohun at the National Institute for Medical
Research, Mill Hill, London. The ducts were also
reconstructed by JJG in the public domain 3D recon-
struction program ‘Reconstruct’ [13] using a subset of
170 sections from the EFIC image stack. A total of 19
ducts were traced and visualised.
Results
Visualising whole-breast ductal anatomy
In the case of the ‘subgross’ breast data the largest duct
system was encoded in its entirety by manual extraction
of 440 straight line sections, with some pruning of ter-
minal branches. The CVAPP model created in this
manner is shown in Figure 1. This establishes that it is
possible to represent the 3D branching of an entire duct
system using data extracted from thick serial sections,
but manual data extraction was laborious and error-
prone, requiring extensive checking. The absence of any

Figure 1. Branching skeleton of a complete duct system in an adult
human breast. Data from subgross slices visualised in CVAPP. This is
a ‘snapshot’ view of a 3D model.
closed ‘cycles’ in this particular duct system can be ob-
served by rotating the model in CVAPP, but cannot
easily be appreciated from this single 2D projection.
Episcopic imaging of nipple ducts
The stack of 1100 images representing the nipple
revealed the lumens of the major ducts as absence of signal
(dark) against the bright autofluorescent background of
the connective tissue of the nipple (Figure 2). There was
variation in diameter between ducts and along their
length. Duct profiles were generally rounded except
furthest from the nipple surface where they began to
adopt the convoluted outlines characteristic of the cen-
tral breast ducts. Fine detail was visible. Viewing images
in sequence as a movie showed the course of a particular
duct through the tissue. As well as viewing data in the xy
plane in which it was collected it was also possible to
view data in the xz and yz planes also (Figure 3). Fine
detail was visible. As the surface was approached, the
Figure 2. Cross section in the x y plane through the ducts of a mas-
tectomy nipple about 1 mm below the skin surface, imaged by auto-
fluorescence EFIC. This is one image out of a stack of 1100. No prior
tissue staining has been used, so cytological detail is not captured by
this approach.
Human breast duct anatomy 287
duct lumens became smaller. Some would eventually
disappear altogether but others could be followed to the
surface, supporting the distinction between ‘type A’ and
‘type B’ nipple ducts proposed by Moffat and Going
[10]. To visualise this anatomy in greater detail, all
identifiable ducts (N=19) were traced individually in
every sixth section in the image stack (N=170) and
displayed in the ‘Reconstruct’ program. An image of all
the ducts (Figure 4) confirms that some can be traced
onto the skin surface while others cannot. Figure 5
contrasts two of these ducts in a closer view. EFIC
images did not, however, have sufficient resolution to
discriminate reliably between keratin plugging and duct
termination beneath the skin surface in the case of ducts
which could not be traced onto the skin surface.
Discussion
Duct anatomy in the nipple
The ducts in the nipple present some paradoxical fea-
tures. There are ducts which can be cannulated from the
outside but do not appear to go anywhere in the breast
[14]. There are sizable ducts in the nipple duct bundle
which cannot be traced onto the skin surface [10]. There
are many more ducts in the nipple duct bundle than
appear to yield milk or can be cannulated, and indi-
Figure 3. EFIC reconstructed in the y z plane. Images 816, 820, 824,
828 and 832. These five closely spaced images of EFIC data show two
ducts which communicate with the skin surface (at the top in each
image). Horizontal banding in these images is an illumination artifact.
288 JJ Going et al.
Figure 4. Three D reconstruction of ducts in the mastectomy nipple.
The inset shows how the 3D reconstruction relates to the intact nipple
(of a normal volunteer). Some ducts clearly reach the skin surface but
others do not identifiably do so. This may either be a consequence of
the lumen being blocked by keratin, or the duct may truly not open
onto the nipple surface.
vidual duct systems vary greatly in their extent [10].
Whether these anatomical differences imply differences
in function or susceptibility to disease remains un-
known, but studies based on the approaches described
in this paper may contribute to their resolution.
Love and Barsky [14] found that on average 5–9
ducts yield milk in the lactating breast, a figure com-
parable to the 9 central ducts identified by ultrasound
in lactating women [15]. These numbers are inconsis-
tent with the well-documented median of 27 ducts at
the base of the papilla in mastectomy breasts [10].
Teboul and Halliwell [16] have suggested that this large
number is a consequence of duct branching within the
nipple. An alternative model, based on 3D recon-
struction of a mastectomy nipple, postulates two dif-
ferent populations of ducts in the nipple: a minority
corresponding to the 5–9 ducts yielding milk during
lactation, and a larger number of ducts tapering to an
origin from skin appendages, but not actually opening
on the skin surface [10]. In this model there is no duct
branching in the nipple. The EFIC data reported here
did not show any branching in the first 2 mm or so
below the skin surface, which does not exclude
branching in the papilla at a deeper level, but EFIC is
Figure 5. Two individual ducts seen in close-up. On the left a duct
terminates before it reaches the skin surface; to the right a duct is
clearly traced in continuity onto the skin surface.
confirmed as an effective tool for examining this
question, although it may not always discriminate
between disappearance of lumens due to keratin plug-
ging (in the case of type A ducts) or decreasing to very
small calibre (in the case of type B ducts). A fluores-
cent nuclear stain revealing duct epithelium in EFIC
images should clarify this issue.
Duct anatomy in the body of the breast
Many secondary accounts of breast anatomy describe
radial ‘lobes’ approximately equal in size, like segments
of an orange or similar fruit. Primary data supporting
this model is rarely quoted and Astley Cooper himself
presented evidence against it [1]. Cooper injected dif-
ferent-coloured masses into individual duct systems in
the breasts of women dying in established lactation.
Dissection of individual duct systems showed that dif-
ferent duct systems vary greatly in size and may lie over
or under one another, rather than adopting a strictly
radial configuration. Cooper likens their intertwinings
to the roots of a tree.
Duct injection studies in mastectomy breasts may be
compromised by extravasations, blockages, leakage
from cut ends, and the difficulty of injecting every duct
system. It seems unlikely that the small-calibre ‘type B’
ducts described by Going and Moffat would readily be
injected. In ‘subgross’ thick section studies the ramifi-
cations of a duct system can be followed wherever they

go, and with patience it is possible to visualise all ducts
and their branching in a complete breast.
Anastamoses
Duct anastamoses allowing abnormally proliferating
and motile cells to escape from one duct system into
another could explain the very extensive DCIS some-
times observed. Anastamoses could also be significant
for intraduct approaches to breast diagnosis and treat-
ment: without anastamoses there is only one path
between a duct opening on the apex of the nipple and
any parenchymal unit derived from that central duct.
Anastamoses could provide alternate pathways. Ohtake
has reported anastamoses [17,18] but Going and Moffat
[10] did not find any in the breast they studied inten-
sively; Love and Barsky [14] did not find evidence of
anastamoses in several different studies, including a
review of an extensive ductography archive; and Cooper
himself says that in all the dissections he made, only one
anastamosis was ever found. This question remains
undecided. Experience of 3D duct modelling suggests
that great care is required to avoid errors in duct tracing,
which could create a false impression that anastamoses
exist when, in fact, they do not.
Data structures representing duct branching
The necessity imposed by the CVAPP data structure to
identify ‘parent’/‘child’ relationships explicitly for every
pair of line segments complicated duct tracing from
subgross material. Although a compact representation of
the data, it required endless cross-reference between
adjacent thick sections to follow ducts correctly, and
complexity of duct branching limited the manual duct
tracing studies of Going and Moffat to one case. It would
be a great advantage if structures present in a thick sec-
tion could be encoded without having to look at adjacent
sections. Working on 2D photographs, one only knows
that a duct leaves a thick section heading ‘up’ into section
n)1 because a corresponding duct appears in that section
and not in section n+l, so one must look at the adjacent
sections to find out which is the case. Stereoscopic images
would show whether a duct exits slice n on the side of
slice n)1 or slice n+l. Instead of the parent/child data
structure, each duct element could be described by the
<x><y><z> coordinates of both ends. Once this
had been achieved for all duct elements in every thick
section ducts leaving section ‘n’ would need to be mat-
ched to ducts entering sections n)1 or n+1 in the rele-
vant plane. In this data structure, separate 2D ‘point
clouds’ correspond to the ‘down’ side of slice n and the
‘up’ side of slice n+l. If section preparation has caused
little distortion, translation and rotation in the z plane
will bring the two point clouds into registration and al-
low matching of the corresponding points. This is a well
understood problem in the computing sciences, with a
number of algorithmic solutions [19–21]. Iteration of this
Human breast duct anatomy 289
procedure from one plane to the next would bring the
whole stack into alignment and from this aligned data
individual duct systems and their branching would be
extractable and modellable.
Subgross sections are suitable for subsequent histo-
logical processing, and the digital model could be used
to show whether histological abnormalities involve sin-
gle or multiple duct systems, a point of interest in rela-
tion to the development of mammary pathology and
neoplasia, and to map molecular as well as histological
changes into their whole-breast developmental/mor-
phological context.
Optimising for subsequent molecular analysis
To optimise the scope of subsequent molecular analysis,
some drawbacks of classical subgross methods may need
to be circumvented. Prolonged fixation in formaldehyde
before cutting thick sections would be impossible to
integrate with tissue processing in a diagnostic service
laboratory, and damages tissue antigens and nucleic
acids, compromising immunohistochemistry and other
analyses such as PCR. Overnight cooling on ice may
allow slicing at appropriate thickness prior to fixation
[4], so the ‘subgross’ methodology should be adaptable
for later molecular analysis. Fixation in 70% ethanol
may be more suitable for subsequent molecular analysis
than formaldehyde [22], and fluorescent dyes like the
dimeric cyanines which become intensely fluorescent
when bound to DNA [23] may have advantages over
haematoxylin and alum carmine.
‘Field change’ and cancer risk in human breast
A motive for seeking to understand whole-breast duct
anatomy is to understand breast cancer origins, and to
investigate possible patterns of ‘field change’ in breast.
The emergence of more than one malignancy in a tissue
field (e.g. oesophageal squamous cancer in a patient who
has had a previous head and neck cancer) may follow
exposure to common aetiological factors. Some (if not
all; [24]) bilateral breast cancer is likely to be explained
on such a basis. But other kinds of ‘field change’ may be
possible. A cell that acquired a growth-promoting
mutation might establish a large clone of descendents,
which might not be histologically detectable if there was
no morphological phenotype. Such a cell arising before
thelarche could be the founder of an extensive duct sys-
tem in its own right; X inactivation ‘patch size’ in normal
breast is at least as big as a lobule [25,26], but could be
larger; information on this point could show how widely
a mutation prior to thelarche might be dispersed within
developing breast. Such a cell arising after thelarche
could give rise to daughter cells replacing the normal
population, increasing cancer risk. Either mechanism
could create a ‘sick lobe’ at increased risk of neoplastic
transformation, so the lobar architecture of the breast is
potentially relevant to mammary carcinogenesis and also
290 JJ Going et al.
diagnosis and treatment of preinvasive changes in breast.
The case for the ‘sick lobe’ breast cancer hypothesis has
recently been argued by Tibor Tot [27].
‘Sick lobe’ formulations have a long history, being
found, for example, in Cheatle and Cutler [28] but the
inaccessible character of the mammary lobe has made
investigation difficult. An alternative to the ‘sick lobe’
hypothesis would be a ‘distributed risk’ model without
major heterogeneity of risk between lobes. These alter-
natives have practical implications for the evaluation of
the cancer risk in a breast by analysis of nipple aspirate
fluid, random peri-areolar fine needle aspiration, or duct
lavage specimens [2,29]. Under a pure ‘distributed risk’
model, cells from any duct system would be representa-
tive of the cells of the whole breast, whereas a pure ‘sick
lobe’ model would require analysis of cells from that
particular lobe to evaluate the cancer risk to that breast.
The truth probably lies between pure ‘dispersed risk’
and ‘sick lobe’ models. ADH in a biopsy predicts risks
to both breasts, and cancer in one breast puts the other
at risk; both are therefore markers of dispersed risk;
DCIS also predicts disease at the same site and appears
therefore to be a more committed progenitor of invasive
disease. The degree to which risk can be localised will
influence the prospects for intraductal diagnostics and
therapy [30].
To be effective, intraductal therapeutics would either
have to be delivered to all or most of the mammary
epithelium (to deal with a dispersed risk) or specifically
to a lobe in which a particular risk had been identified.
Improved knowledge of the ductal lobar anatomy would
be relevant to the treatment planning under both
scenarios. Risk reducing epithelial ablation would not
need to be 100% effective to compete with surgical risk-
reducing mastectomy. If a relatively small number of
ducts do actually communicate with a majority of the
breast tissue, it might be feasible.
Electronic Supplementary Material Supplementary
material is available for this article at http://dx.doi.org/
10.1007/s10549-005-9122-7
Acknowledgements
This study was partly supported by the University of
Glasgow.
References
1. Cooper AP: On the Anatomy of the Breast. Longman, Orme,
Green, Brown and Longmans, London, 1840
2. Fabian CJ, Kimler BF, Mayo MS, Khan SA: Breast-tissue sam-
pling for risk assessment and prevention. Endocr Relat Cancer 12:
185–213, 2005
3. King BL, Love SM, Rochman S, Kim JA: The fourth interna-
tional symposium on the intraductal approach to breast cancer,
Santa Barbara, California, 10–13 March 2005. Breast Cancer Res
7: 198–204, 2005
4. Egan RL, Ellis JT, Powell RW: Team approach the study of
diseases of the breast. Cancer 23: 847–854, 1969
5. Jensen HM, Rice JR, Wellings SR: Preneoplastic lesions in the
human breast. Science 191: 295–297, 1976
6. Marcum RG, Wellings SR: Subgross pathology of the human
breast: method and initial observations. J Natl Cancer Inst 42:
115–121, 1969
7. Wellings SR, Jensen HM, Marcum RG: An atlas of subgross
pathology of the human breast with special reference to possible
precancerous lesions. J Natl Cancer Inst 55: 231–273, 1975
8. Purnell DM, Stowers DJ: Mammary gland hyperplastic alveolar
noduligenesis in the LEW/Mai rat in relation to mitotic activity,
estrous cycle status, and age. J Natl Cancer Inst 61: 1339–1346,
1978
9. Weninger WJ, Mohun T: Phenotyping transgenic embryos: a rapid
3-D screening method based on episcopic fluorescence image
capturing. Nat Genet 30: 59–65, 2002
10. Going JJ, Moffat DF: Escaping from Flatland: clinical and bio-
logical aspects of human mammary duct anatomy in three
dimensions. J Pathol 203: 538–544, 2004
11. Moffat DF, Going JJ: Three dimensional anatomy of complete
duct systems in human breast: pathological and developmental
implications. J Clin Pathol 49: 48–52, 1996
12. Cannon RC, Turner DA, Pyapali GK, Wheal HV: An on-line
archive of reconstructed hippocampal neurons. J Neurosci
Methods 84: 49–54, 1998
13. Fiala JC: Reconstruct: a free editor for serial section microscopy.
J Microsc 218: 52–61, 2005
14. Love SM, Barsky SH: Anatomy of the nipple and breast ducts
revisited. Cancer 101: 1947–1957, 2004
15. Ramsay DT, Kent JC, Hartmann RA, Hartmann PE: Anatomy of
the lactating human breast redefined with ultrasound imaging.
J Anat 206: 525–534, 2005
16. Teboul MHT: Atlas of Ultrasound and Ductal Echography of the
Breast: The Introduction of Anatomic Intelligence Into Breast
Imaging. Oxford, Blackwell, 1995
17. Ohtake T, Abe R, Kimijima I, Fukushima T, Tsuchiya A, Hoshi
K, Wakasa H: Intraductal extension of primary invasive breast
carcinoma treated by breast-conservative surgery. Computer gra-
phic three-dimensional reconstruction of the mammary duct-lob-
ular systems. Cancer 76: 32–45, 1995
18. Ohtake T, Kimijima I, Fukushima T, Yasuda M, Sekikawa K,
Takenoshita S, Abe R: Computer-assisted complete three-dimen-
sional reconstruction of the mammary ductal/lobular systems:
implications of ductal anastomoses for breast-conserving surgery.
Cancer 91: 2263–2272, 2001
19. Besl PJ, McKay ND: A method for registration of 3D shapes.
IEEE Trans Pattern Anal Mach Intell 14: 239–256, 1992
20. Griffin P, Alexopoulos C: Point pattern matching using centroid
bounding. IEEE Trans Syst Man Cybern 19: 1274–1276, 1989
21. Murtagh F: A new approach to point-pattern matching. Publ
Astron Soc Pac 104: 301–307, 1992
22. Gillespie JW, Best CJ, Bichsel VE, Cole KA, Greenhut SF, Hewitt
SM, Ahram M, Gathright YB, Merino MJ, Strausberg RL,
Epstein JI, Hamilton SR, Gannot G, Baibakova GV, Calvert VS,
Flaig MJ, Chuaqui RF, Herring JC, Pfeifer J, Petricoin EF,
Linehan WM, Duray PH, Bova GS, Emmert-Buck MR: Evalua-
tion of non-formalin tissue fixation for molecular profiling studies.
Am J Pathol 160: 449–457, 2002
23. Bink K, Walch A, Feuchtinger A, Eisenmann H, Hutzler P, Hofler
H, Werner M: TO-PRO-3 is an optimal fluorescent dye for nuclear
counterstaining in dual-colour FISH on paraffin sections. Histo-
chem Cell Biol 115: 293–299, 2001
24. Schlechter BL, Yang Q, Larson PS, Golubeva A, Blanchard RA,
de las Morenas A, Rosenberg CL: Quantitative DNA finger-
printing may distinguish new primary breast cancer from disease
recurrence. J Clin Oncol 22: 1830–1838, 2004
25. Novelli M, Cossu A, Oukrif D, Quaglia A, Lakhani S, Poulsom R,
Sasieni P, Carta P, Contini M, Pasca A, Palmieri G, Bodmer W,
Tanda F, Wright N: X-inactivation patch size in human female
tissue confounds the assessment of tumor clonality. Proc Natl
Acad Sci USA 100: 3311–3314, 2003

26. Tsai YC, Lu Y, Nichols PW, Zlotnikov G, Jones PA, Smith HS:
Contiguous patches of normal human mammary epithelium de-
rived from a single stem cell: implications for breast carcinogene-
sis. Cancer Res 56: 402–404, 1996
27. Tot T: DCIS, cytokeratins, and the theory of the sick lobe. Vir-
chows Arch 447: 1–8, 2005
28. Cheatle GL, Cutler M: Tumours of the Breast. Edward Arnold,
London, 1931
29. Mokbel K, Escobar PF, Matsunaga T: Mammary ductoscopy:
current status and future prospects. Eur J Surg Oncol 31: 3–8, 2005
Human breast duct anatomy 291
30. King BL, Love SM, Rochman S, Kim JA: The Fourth Interna-
tional Symposium on the Intraductal Approach to Breast Cancer,
Santa Barbara, California, 10–13 March 2005. Breast Cancer Res
7: 198–204, 2005
Address for offprints and correspondence: James J. Going, Division of
Cancer Sciences and Molecular Pathology, University of Glasgow,
Glasgow, Scotland, UK; Tel.: +141-211-5110; Fax: +141-211-2442;
E-mail: going@udcf.gla.ac.uk